Research & Innovation
Seeing Speech Animates Faces by Modelling Three Motions: Spreading, Opening and Protrusion
Speech-driven 3D facial animation has better vertex-level reconstruction than ever and still struggles with speech-consistent visible articulation, because the mapping from acoustics to motion is one-to-many. Seeing Speech, posted September 24, models visible speech as three directional articulatory motions — spreading, opening and protrusion — composed into surface-consistent facial motion.